International Journal of Engineering and Advanced Technology (IJEAT)

ISSN: 2249 – 8958, Volume-8 Issue-6, August 2019

Evaluation of Machine Learning Algorithms for

Crop Yield Prediction
Renuka, Sujata Terdal

 dimensions. Technically, data mining is finding correlations

Abstract: Agriculture plays a significant role in the growth of the or patterns of fields in large relational databases. All of these
national economy. It relay on weather and other environmental data can provide information between models, connections,
aspects. Some of the factors on which agriculture is dependent are or relationships. Knowledge can be transformed into
Soil, climate, flooding, fertilizers, temperature, precipitation,
historical patterns and knowledge of future trends. For
crops, insecticides and herb. The crop yield is dependent on these
factors and hence difficult to predict. To know the status of crop example, a survey of agricultural products helps farmers to
production, in this work we perform descriptive study on suggest and prevent future crop losses.
agricultural data using various machine learning techniques. Many research have been conducted to develop an efficient
Crop yield estimates include estimating crop yields from available method for yield prediction but focus have been always on
historical data such as precipitation data, soil data, and historic statistical techniques and not much has been done in machine
crop yields. This prediction will help farmers to predict crop yield learning approach. The crop production depends on various
before farming. Here we are utilizing three datasets like as clay factors [3] which change with every square meter and depends
dataset, precipitation dataset, and production dataset of
on:
Karnataka state, then we structure an assembled data sets and on
this dataset we employ three different algorithms to get the 1. Geography of region,
genuine assessed yield and the precision of three different 2. Weather (Temprature, humidity ,percipitation),
methods. K-Nearest Neighbor(KNN), Support Vector 3. Soil type (saline, alkaline,sodic,non-alkaline),
Machine(SVM), and Decision tree algorithms are applied on the 4..Soilcomposition(ph,N,P,K,EC,OC,Zn,F).
training dataset and are tested with the test dataset, and the
implementation of these algorithms is done using python Various subsets of these parameters are used in different
programming and spyder tool. The performance comparison of prediction models for various crops. Prediction models are
algorithms is shown using mean absolute error, cross validation
essentially two main types.
and accuracy and it is found that Decision tree is giving accuracy
of 99% with very less mean square error(MSE). The proposed 1. Statistical models, which use a single prediction function
model can exhibit the precise expense of assessed crop yield and it that includes all sample spaces.
is mark like as LOW, MID, and HIGH. 2. Machine learning technology, a new technology for
knowledge search that connects input and output variable
Keywords: Descriptive analytics, Agriculture, Machine learning, models.
K-Nearest Neighbour, Support Vector Machine, Decision tree. Machine learning has the ability to learn the machine without
defined computer programming, so it improves machine
I. INTRODUCTION performance by detecting and characterizing the consistency
Agriculture is one of the important industrial sectors in India and pattern of drive data. Machine learning can be classified
and the country's economy relies heavily on it for the into three categories according to the learning method
sustainability of its rural areas. Due to some factors [1] such –Supervised learning, Unsupervised learning and
as climate change, unplanned rainfall, falling water levels, Reinforcement learning. In our project we are making use of
excessive use of pesticides etc., the level of agricultural supervised learning algorithms to predict crop yield. This type
production in India is declining. Most farmers do not achieve of algorithms helps to build most accurate and effective model
expected crop yield for a variety of reasons. To understand because here, the learning data comes with labels or desired
production levels, yield prediction is carried out which outputs and the objective is to find a general rule of mapping
involves predicting the yield of the crop based on the existing input to output. It involves building a machine learning model
data. Previously, crop yield estimates were based on farmer's that is based on labeled samples.
specific crops and cultivation experience. The proposed system analyzes the application of supervised
. machine learning approaches in forecasting sugar cane yield
There are many ways to enhance and improve crop yield and of Karnataka region. The dataset is collected from several
quality. Data mining techniques are also helpful for predicting online sources such as Kaggle.com and data.gov.in, it consist
crop yields. In general, data mining analyzes knowledge from of rainfall dataset, yield dataset and soil dataset. The SVM,
various approaches and summarizes it as profitable KNN and Decision trees have been used for yield prediction.
information. Data Mining software [2] is an analytical tool This paper is organized as follows : Section II presents the
that allows users to classify and summarize identified related work ,whereas the proposed method is discussed in
relationships as well as analyze data at various angles or Section III.

Revised Manuscript Received on July 22, 2019.

Renuka, Computer Science and Engineering, PDACE, Kalaburagi, India.
Dr. Sujata Terdal, Computer Science and Engineering, PDACE,
Kalaburagi, India.

Published By:
Retrieval Number F8640088619/2019©BEIESP Blue Eyes Intelligence Engineering
4082 & Sciences Publication
DOI: 10.35940/ijeat.F8640.088619
 Evaluation of Machine Learning Algorithms for Crop Yield Prediction

Then, the experimental results on agricultural data are The dataset is processed through WEKA tool to build the set
discussed in Section IV. Finally, conclusion is given in of rules on the current dataset. The results were generated in
Section V. python by using SVM algorithm. In [15] based on the C4.5
algorithm, decision tree and decision rules have been
II. RELATED WORK developed, in their study they have developed a website called
Crop Advisor: This an interactive website for discovering the
Machine learning in Agriculture is a Novel field, a great deal
affect of weather and crop production by using C4.5
of work has been done in field of Agriculture utilizing
algorithm[15]. This gives the idea of how different climatic
Machine learning. There are diverse guaging philosophies
parameters impact the growth of the crop. The selections were
created and assessed by the specialists everywhere throughout
made based on the area under the chosen crop. The
the world in the field of farming or related sciences.
information regarding the associated years climatic
Agricultural scientists in Pakistan have demonstrated that
parameters like rainfall, high and low temperature, wet day
endeavors of harvest yield amplification through expert
frequency where collected. The id3 algorithm [16] were
pesticide state strategies have prompted a hazardously high
developed to get good quality and improved Tomato crop
pesticide use. These examinations have revealed negative
yield which is implemented in PHP platform and uses csv as
relationship between's pesticide use and harvest yield [5]. In
data sets. The features used in this study include area,
their investigation they have explained that how data mining
production of tomato crop, temperature and humidity.
incorporated farming information including irritation
exploring, pesticide utilization and meteorological
III. PROPOSED SYSTEM
information are helpful for streamlining of pesticide use.
Topical data identified with agribusiness which has spatial In the proposed system, we use supervised learning to form a
properties was accounted for in one of the study[6]. model, which provides predicted cost of crop yield and
Their research went for perceiving patterns in corresponding production order. The proposed system is
farming creation with references to the accessibility described in following stages such as dataset collection,
of information assets. K-means method turned preprocessing step, feature selection and applying machine
into applied to carry out gauges of the contamination in learning modules as shown in figure 1.
the air[7],the k- nearest neighbor become connected for
mimicking day by day precipitations A. Dataset Collection: Data is collected from a variety of
and other climate elements [8],and numerous ability changes sources and prepared for data sets. And this data is used for
of the weather situations are descriptive analysis. Data is available from several online
dissected utilizing SVM[9]. Statistics mining techniques are abstract sources such as Kaggle.com and data.gov.in. We will
often used to have a look at soil qualities. As example, the k- use an annual summary of crops for at least 10 years. The data
means method is used for segmenting soils in mixture with sets used in this paper are soil dataset, rainfall dataset and
GPS-based technology [10]. A decision tree classifier for crop yield data.
agriculture information turned into proposed [11].This new
classifier uses new facts expression and can address each B. Preprocessing step: This step is a very important step in
entire records and in entire records. Inside the test,10-fold machine learning. Preprocessing consists of inserting the
cross validation technique is used to check the dataset, missing values, the appropriate data range, and extracting the
horse-colic dataset and soybean dataset. Their results showed functionality. The kind of the dataset is critical to the analysis
the proposed selection tree is capable of classifying all styles process. In this paper we have used isnull() method for
of agriculture records. A yield prediction version turned into checking null values and lable Encoder() for converting the
proposed in one of the take a look at [12] which makes use of categorical data into numerical data.
data mining techniques for category and prediction. This
model worked on enter parameters crop name, land location, C. Feature Selection: Feature extraction should simplify the
soil type, soil ph, pest information, climate, water stage, seed amount of data involved to represent a large data set. The soil
type and this model anticipated the plant boom and plant and crop characteristics extracted from the pre-treatment
diseases and therefore enabled to select the nice crop based on phase constitute the final set of training. These characteristics
climate information and required parameters. include the physical and chemical properties of the soil. Here,
There are few research works about sugarcane yield we have used Random Forest Classifier() method for feature
prediction which can be associated with our work. Sugarcane selection. This method selects the features based on the
yield prediction technique with use of Random forest [13] entropy value i.e., the attribute which is having more entropy
became proposed in one of the survey, the features used in this value is selected as important feature for yield prediction.
study consist of biomass index, climate statistics (e.g.,
rainfall) and yields from previous years. Two predictive tasks
are provided in [13]: (i) the category problem for predicting
whether or not the yield can be above or underneath the found
median yield, and (ii) the regression hassle for predicting the
yield estimates in two distinct time intervals. In addition,
support vector system[14] for rice crop yield prediction
become proposed, the dataset used in this method are
precipitation ,minimum, maximum and common temperature,
place, evapotranspiration and manufacturing. The sequential
minimal optimization classifier is implemented on the dataset.

Published By:
Retrieval Number F8640088619/2019©BEIESP Blue Eyes Intelligence Engineering
4083 & Sciences Publication
DOI: 10.35940/ijeat.F8640.088619
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-8 Issue-6, August 2019

Flow chart:

Fig.1 System Architecture

D. Split the Dataset into Train and Test Set: This step
includes training and testing of input data. The loaded Fig.2 Flow chart for KNN algorithm
data is divided into two sets, such as training data and test ii. Support Vector Machines
data, with a division ratio of 80% or 20%, such as 0.8 or
0.2. In a learning set, a classifier is used to form the SVM divides the given data into decision surface. Decision
available input data. In this step, create the classifier's surface is further divide the data into hyper plane of two
support data and preconceptions to approximate and classes. Training points defines the supporting vector which
classify the function. During the test phase, the data is defines the hyper plane. Probably, a hyper plane with the
tested.The final data is formed during preprocessing and greatest distance to the closest learning data point typically
is processed by the machine learning module. has better margins and larger errors because of the larger
margins, the generalization of classifiers is weak. The flow
E. Applying Machine Learning modules: In our project chart for SVM is given in the figure 3, it shows the steps
we have used three different supervised machine learning involved in SVM algorithm.
algorithms for crop yield prediction which is given as
follows Flow Chart:
i. KNN Algorithm

KNN is a nonparametric supervised learning technique that

uses training sets to segment data points into given categories.
In simple classifications, the word collects information from
all educational cases and similarities based on the new case.
Look at the training for the most similar (neighbor) K cases
and predict the new instance (x) by summarizing the output
variables for these K cases. Classification is the class value
mode (or most commonly). A flow diagram of the KNN
algorithm is shown in Figure 2.

Fig.3 Flow chart for SVM algorithm

Published By:
Retrieval Number F8640088619/2019©BEIESP Blue Eyes Intelligence Engineering
4084 & Sciences Publication
DOI: 10.35940/ijeat.F8640.088619
 Evaluation of Machine Learning Algorithms for Crop Yield Prediction

iii. Decision tree different parameters set for these algorithms were mean
Decision tree is a predictive model which works by checking squared error, accuracy and cross validation which are used to
condition at every level of the tree and proceeds towards estimate the efficiency of the method as shown in the table 1.
bottom of the tree where various decisions are listed. The
condition depends on the application and the outcome might The formula for calculating Accuracy and Mean Squared
be in terms of decision. There are various types of Decision Error(MSE) are:
tree algorithms such as C4.5,CART and ID3 algorithm. In this
paper we are using ID3 algorithm for model building. The
factors like Rainfall, soil type and crop yield can be used as
input to Prediction algorithm and the resultant will be in terms
of the prediction of productivity.
The ID3 algorithm [15] starts with the original set S as the
root node. Each iteration of the algorithm finds each unused
attribute in set S and computes the entropy H (S) (or
information gain IG (S)) of this attribute. Then select the Table 1.Comparision table for different parameters
attribute with the smallest entropy value (or the largest
information gain). The set S is then divided into the selected Algorithms Accuracy MSE
attributes. The algorithm is continually reproduced in each Decision tree 99 0.01
subset, considering only those attributes that were not
KNN 58.078 0.603049
previously selected. The figure 4 shows the flow diagram of
SVM_rbf 58.952 0.559123
the ID3 algorithm.
SVM_linear 53.712 6.209897

Machine learning algorithms have been applied individually

using the Cross Validation techniques with 10 folds and
accuracy of prediction has been observed for each of them. In
this paper the accuracy for SVM was calculated for two
different kernels i.e, svm _rbf and svm_linear among these
two RBF kernel was showing more error rate. The decision
tree is giving more accuracy with very less MSE. The table 2
shows cross validation runs for different algorithms.

Table 2. Cross validation runs for different algorithms

Cross Decision KNN SVM_rbf SVM_line

validati tree ar
on
1 0.992375 0.959368 -0.058233 0.9306726
2 0.945654 0.886403 -0.189782 0.8693865
3 0.814398 0.940818 -0.191459 0.9343228
4 0.896650 0.940682 - 0.140031 0.7896232
5 0.875382 0.872272 -0.197912 0.7903085
6 0.976996 0.956903 -0.136351 0.9011202
7 0.942567 0.976868 -0.116380 0.8950682
8 0.944686 0.962245 -0.147855 0.8929110
9 0.909315 0.876680 -0.102809 0.7334802
10 0.949203 0.956820 -0.144903 0.8934893
Fig.4 Flow chart for ID3 algorithm
The following figure 5 and 6 shows comparison graph for
Once the model has been trained efficiently it is tested on the accuracy and mean squared value for KNN, SVM_RBF,
Testing dataset which is different from the Training data in SVM_LINEAR and Decision tree. As shown in the graph the
sample values. accuracy for decision tree is more and also it is showing less
error rate.
IV. DISCUSSION AND RESULTS
This section shows the results obtained after implementation
of the Machine learning algorithms on sugar cane crop data
set of Karnataka state, India. The algorithms KNN, SVM and
decision tree are applied on yield dataset, soil dataset and
rainfall dataset. In this paper, we have used Spyder an open
source, multi-platform integrated development (IDE)
environment for Python scientific programming. The

Published By:
Retrieval Number F8640088619/2019©BEIESP Blue Eyes Intelligence Engineering
4085 & Sciences Publication
DOI: 10.35940/ijeat.F8640.088619
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-8 Issue-6, August 2019

3. Vaneesbeer Singh, Abid Sarwar, Vinod Sharma. “Analysis of soil and

prediction of crop yield (Rice) using Machine Learning approach”,
IJARCS 8 (5), May-June 2017, pp 1254-1259.
4. A.S. Ponraj, Vigneswaran T. “Machine Learning Approach for
Agricultural IoT”,IJRTE, Volume-7, Issue-6, March 2019,pp 383-391.
5. Abdullah, A., Brobst, S, Pervaiz.I., Umer M.,and A.Nisar, “Learning
dynamics of pesticide abuse through data mining”, Proceedings of
Australian Workshop on Data mining and Web Intelligence, New
Zealand, January .2004,pp 63-68.
6. Kiran Mai,C., Murali Krishna, I.V, an A.VenugopalReddy, “ Data
Mining o f Geospatial Database for Agriculture Related Application”,
Proceedings of Map India,New Delhi, 2006,pp 83-96.
7. Jorquera H, Perez R, Cipriano A, Acuna, “Short term forecasting of air
pollution episodes”,In. Zannetti P (eds) Environmental Modeling 4.
G(2001) WITPress, UK,pp 221-237.
8. Rajagopalan B. Lall, “ A k- nearestneighbor daily precipitation and
other weather variables.” U (1999) WatResResearch35(10) :3089 –
3101.
Fig.5 Comparison graph for Accuracy 9. Tripathi S, Srinivas VV, Nanjudiah , “Down scaling of precipitation
for climate change scenarios: a support vector machine approach”, RS
(2006),J. Hydrology 330-337.
10. Verheyen, K., Adrianens, M. Hermy and S.Deckers(2001). High
resolution continuous soil classification using morphological soil
profile descriptions. Geoderma, 101:31-48.
11. Jun Wu, Anastasiya Olesnikova, Chi- Hwa Song, Won Don Lee
(2009).The Development and Application of Decision Tree for
Agriculture Data. IITSI, pp 16-20.
12. R. Sujatha and P. Isakki, "A study on crop yield forecasting using
classification techniques," 2016 International Conference on
Computing Technologies and Intelligent Data Engineering
(ICCTIDE'16), Kovilpatti, 2016, pp. 1-4.
13. Y. Everingham, J. Sexton, D. Skocaj, and G. Inman-Bamber.
“Accurate prediction of sugarcane yield using a random forest
algorithm”, Agronomy for Sustainable Development, vol. 36, no. 2,
2016.
14. N. Gandhi, L. J. Armstrong, O. Petkar and A. K. Tripathy, "Rice crop
Fig.6 Comparison graph for MSE yield prediction in India using support vector machines," 2016 13th
International Joint Conference on Computer Science and Software
Engineering (JCSSE), Khon Kaen, 2016, pp. 1-5.
V. CONLUSION 15. S. Veenadhari, B. Misra and C. Singh, "Machine learning approach for
Different machine learning algorithms have been forecasting crop yield based on climatic parameters," 2014
International Conference on Computer Communication and
implemented on agricultural data to evaluate the best Informatics, Coimbatore, 2014, pp. 1-5.
performing method. In this paper, we used three different 16. CH. Vishnu Vardhan chowdary, Dr.K.Venkataramana, “Tomato Crop
supervised learning algorithms, such as SVM, KNN, and Yield Prediction using ID3”, March 2018,IJIRT Volume 4 Issue 10
decision trees. This study provides information on how to pp,663-62.
apply data analysis to the data set of sugar cane crops. There
are three sets of data: soil data sets, rainfall data sets, and crop AUTHORS PROFILE
yield data sets. This data set consists of a variety of
parameters that are useful for identifying status of crops and Renuka Mtech student in Computer Network
conducting supervisory training on data sets collected from and Engineering at Poojya Doddappa Appa College
of Engineering, Kalaburagi, Karanataka. Her
agriculture domain to divide information into multiple
Research interest is in the area of Computer
classes. This paper shows the comparison of three different Networks and Machine Learning.
algorithms like, decision tree, KNN and SVM. These
algorithms were used to train the 0.8 or 80 percentage of the
input data and are tested with the remaining 0.2 or 20
percentage of test dataset and results of the algorithms were
compared based on accuracy and mean square error. Here, the Dr. Sujata Terdal M.Tech, Ph.D, Associate
decision tree algorithm is giving more accuracy of 99% and Professor, Computer Science and Engineering
also the mean square error for this algorithm is very less. This Department have teaching experience of 23 years.
Areas of research are Mobile Ad Hoc Networks and
system will help to reduce the problems faced by farmers and Wireless Networks. She has published number of
will serve as an intermediary to provide farmers with the research papers in International and National journals
information they need to earn high profits and maximize and conferences.
profits.

REFERENCES
1. Arun Kumar, Naveen Kumar, Vishal Vats.”Efficient crop yield
prediction using machine learning algorithms”, IJRET Volume: 05
Issue: 06, June-2018, pp 3151-3159
2. P.Priya, U.Muthaiah ,M.Balamurugan. “ Predicting yield of the crop
using machine learning Algorithm”, IJESRT et al., 7(4): April-2018,
pp 2277-2284

Published By:
Retrieval Number F8640088619/2019©BEIESP Blue Eyes Intelligence Engineering
4086 & Sciences Publication
DOI: 10.35940/ijeat.F8640.088619